Assessing, Managing, and Forecasting Social-Ecological Systems

# Assessing, Managing, and Forecasting Social-Ecological Systems
### Dan Ovando
### OSU/ODFW Research Seminar
### 2022/05/24

---

# A Quantiative Fisheries Ecologist

> I use methods and ideas from ecology and economics to help understand and manage social-ecological systems.

I study...
  
  - Fisheries assessment and management
  
  - Social-ecological impacts of policies
  
  - Predictive modeling
  
Using ...
 
  - Bayesian methods
 
  - Econometrics
 
  - Simulation modeling
 
  - Open and reproducible science

???

What do I mean by an optimistic skeptic? Social-ecological systems are complicated. This both complicates our ability to clearly understand what is happening in them, and to identify simple solutions to complex problems. Hence, "skeptic": my prior is that uncertainty and complexity are the norm. Optimist though: I believe that science can help us navigate this uncertainty and complexity in a productive way, to help us move slowly towards better understanding and better outcomes.

---

class: center, middle, inverse
### Follow along at [danovando.github.io/osu-seminar/slides](https://danovando.github.io/osu-seminar/slides#1)

---

# Basics of Bayes
--

- Provide **posterior probability distributions** for parameters, informed by **priors** 
  
  - "What was the probability that the protected area increased biomass by more than 10%"
  
- Give us a framework for incorporating **prior knowledge**
  
- Suited to **complex data generating processes**

---

## 1. Assessing Data-Limited Fisheries

## 2. Effects of Protected Areas

## 3. Vision at OSU/ODFW

---

---

> Counting fish is just like counting trees, except they move and are invisible
> <footer>---  John Shepherd (loosely)</footer>

---

# Fisheries Science

1. How many fish **are there**?

2. How many fish **can we catch**?

We answer these using **stock assessments**
  - Fishery dependent & independent data...
  - Confronted with population dynamics

].pull-right[

![:scale 90%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/IMG_1664.JPG)

]

---

# A Tale of Two Fisheries

Stock assessments provide best estimates of the state of fisheries providing <br> ~50% of the world's catch.

.center[![:scale 80%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/two-fisheries.png)]

---

# How do we assess the "unassessed"?

We don't have good **abundance** data for most of the world's fisheries.

We do have data on **catches**

???
mention abundance indices
---

# Can Catches Inform Global Assessments?

- Catch-only models increasingly used to assess and manage fisheries

- Commonly **infer** or **predict** stock status based on attributes of **catch history**

- Performance vigorously debated
  - "The best we can do" vs. "too uncertain to be helpful"

![](data:image/png;base64,#https://media.springernature.com/w300/springer-static/image/art%3A10.1038%2F494303a/MediaObjects/41586_2013_Article_BF494303a_Figa_HTML.jpg?as=webp)

].pull-right[

]
---

# Assessing Global Data-Limited Fisheries
.pull-left[

### Our Questions:

1. **How well do catch-only models work?**
  
  - Spoiler: Not very well.

2. **Can we make them better?**
  
  - Spoiler:  Not without investment in new data

].pull-right[

<br>
<br>
<br>
.center[![:scale 100%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/two-fisheries.png)]

]
---

# A Flexible Assessment Model

Core tool of our research:  [`sraplus`](https://danovando.github.io/sraplus/)

Pella-Tomlinson surplus production model with process error

`$$B_{t + 1} = \left(B_{t} + B_{t}\frac{r}{m - 1}\left(1 - \left(\frac{B_t}{K}\right)^{m- 1}\right) - c_t\right)p_t$$`
`$$log(p_t) \sim {N}\left(\frac{-\sigma_{proc}^2}{2},\sigma_{proc}\right)$$`

Can... 
  - Run as a catch-only model
  
  - Incorporate subjective or empirical priors on stock status
  
  - Fit to abundance index if available

---

# What Can Catches Tell Us?

What do these catch histories suggest about stock status?

---

#  Catches Don't Tell Us (much) About Status

.center[![:scale 75%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/nothing_learned_plot_a.png)]

#  Catches Don't Tell Us (much) About Status

.center[![:scale 75%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/nothing_learned_plot.png)]

# Moving Beyond Catch-Only

- We can't **infer** stock status based on catch alone
  
  - We can predict some...

- There's a lot of data between "catch-only" and "full integrated stock assessment"
  
  - Effort reconstructions
  
  - Strength of fisheries management
  
  - Intensity of trawling
  
  Using `sraplus`, we tested the ability of different combinations of **broadly available** data to say something useful about the status of fisheries

---

# Example: Swept Area Ratio

Places with heavy trawling are likely to have higher fishing mortality

# Translating Swept Area Ratio

The **posterior prediction** based on trawling intensity becomes **prior** for `sraplus`

---

# Additional Data Helps Individual Fisheries

--
.center[![:scale 80%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/ex_scatter.png)]

???

<small> CMSY is popular catch-only method</small>

<small> SAR & FMI augments catches with local data on trawl fishing and strength of management </small>
---

???
.right[[Ovando *et al.* (2021)](https://onlinelibrary.wiley.com/doi/abs/10.1111/faf.12593)]

---

# Little Improvement from <br> Regional Application

- On average, catch-only 25% better than a random guess
  
  - But, **wrong status 57%** of the time...

- Augmenting with additional data helps some
  
  - Modest reductions in error and bias
  
  - Better at fishing mortality rates

- Estimates have substantial **error** and **bias**
  
  - Our hope might be uncertain but unbiased...

---

# How Can We Improve?

# Assessing Data-Limited Fisheries

- Can't let perfect be the enemy of good enough

- Important to know how imperfect
  
  - Catches don't really update priors
  
- Local data on institutions and lengths could help

].pull-right[

]
---

---

# Leaving the Walled Garden

- Protected areas asked to play a growing role in **ecosystem-based management**
  
  - Protect 30% of land and water in the coming decade.
  
- Early protected areas refuges from the outside world
  
- Protection has wider impacts
  
  - "spillover" of adult or larval organisms
  
  - Displacement of human activities

].pull-right[

]

---

# MPAs - The Evidence

<br>
<br>
<br>
> The jury is in on marine reserves: They work.

> Research has repeatedly shown that fish numbers quickly climb following well-enforced fishing bans, 
> creating tangible benefits for fishers who work the surrounding waters.

.right[[NPR 2018](https://www.npr.org/sections/thesalt/2018/09/14/647441547/could-a-ban-on-fishing-in-international-waters-become-a-reality)]

---

# MPAs - The Evidence

- MPAs seem simple: 
  - *Less fishing, more fish*

- Ample evidence of **more everything** inside MPAs than outside

- Problem solved?

- Few marine populations contained entirely in MPAs

- What do we do when the **treatment** affects the **control**

].pull-right[

]

---

# Population-Level Effects of MPAs

???
Relatively sedentary adults, widely dispersed larvae with post-settlement density dependence

---

# Population-Level Effects of MPAs

.center[![:scale 70%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/fleet_anim.gif)]

???
Relatively sedentary adults, widely dispersed larvae with post-settlement density dependence

---

# Case Study: Channel Islands MPAs

---

# What Happend in the Channel Islands?

Trend for species targeted by fishing... Looks promising!

.center[![:scale 80%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/raw_targ_trend.png)]

---

# What Happend in the Channel Islands?

But wait...

.center[![:scale 80%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/raw_trend.png)]

---

# Estimating Population-Level Effects

---

# Estimating Population-Level Effects

---

# Estimating Population-Level Effects

---

# Estimating Population-Level Effects

### General form

`$$(log(D_{MPA=1, T = 1})-log(D_{MPA=0, T = 1}))-(log(D_{MPA = 1,T = 0})-log(D_{MPA =0, T = 0}))$$`

### Bayesian Gamma GLM

`$$d_{i} \sim Gamma(e^{\beta_0 + \beta_1T_{i} +  \beta_2MPA_{i} + \color{red}{\beta_{3}}T_iMPA_i + \mathbf{B^cX_i} + \mathbf{B^sS_i}},shape)$$`

### Hierarchical clustering of sites by island
`$$\mathbf{B^s} \sim N(\beta_r,\sigma_r)$$`

---

# Population-Level Effects of MPAs

.center[![:scale 82%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/a_priori_plot.png)]

---

# Population-Level Effects of MPAs

.center[![:scale 82%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/mpa_effect_plot.png)]

---

## Effects of Protected Areas

- Protected areas can support ecosystem-based management

- Critical to consider effects **inside** and **outside**

-  **Population** effects may be smaller and harder to find than conventionally thought

- Does not mean there are not benefits
  - But may impede adaptive management
  
- Communities should have a clear understanding of expected effects of protected areas

].pull-right[

]

---

# Vision at OSU/ODFW

---

### A welcoming hub for interdisciplinary research on the <br> functioning and management of social-ecological systems

---

# Reducing Bycatch

- Bycatch remains a problem even in otherwise well-managed fisheries
  - Sharks, turtles, marine mammals

- Static spatial protections may struggle in a changing climate

- Goal at OSU/ODFW: **Theory and evidence for reducing bycatch of threatened species**
  - [`marlin`](https://danovando.github.io/marlin/)
  - Practical applications of dynamic ocean management

].pull-right[

]

---

# `marlin`

---

# West Coast Groundfish

- Groundfish serve critical ecological, cultural, and economic roles

- Range of assessment and management challenges
  
  - diverse life history, data availability, fishery dynamics
  
- Areas of research at OSU/ODFW

- Spatial assessment and management
  
  - Economic indicators in stock assessment
  
  - Modern monitoring tools

].pull-right[

]

---

# West Coast Groundfish

![](data:image/png;base64,#https://danovando.github.io/dissertation/thesis_files/figure-html/cs2-1.png)

---

# Forecasting Social-Ecological Systems

- Quantitative ecology often focuses on explaining what we **observe**
  - Good identification but low `$R^2$`

-  We also need to consider how well we can **forecast**

- Particularly as we plan ecosystem management under a changing climate

- Goal at OSU/ODFW: **Collaborative research forecasting social-ecological systems**
].pull-right[

]
---

# Forecasting Social-Ecological Systems

---

# Philosophy of Healthy Science

Social-ecological problems are **complex**

**Solving** them requires
  
  - Open and reproducible science
  
  - Interdisciplinary collaboration
  
  - Diverse teams

.center[**I'm excited to help support sustainable fisheries and healthy ecosystems as part of OSU/ODFW community**]

] .pull-right[

]

---

# Questions?

**slides:** <br>

[danovando.github.io/osu-seminar/slides](https://danovando.github.io/osu-seminar/slides)

**email:** danovan@uw.edu

**website:** [danovando.com](https://www.weirdfishes.blog/)

**twitter:** [@danovand0](https://twitter.com/DanOvand0)

#### Funding
  
  - Assessment: Food and Agriculture Organization of the United Nations
  
  - Protected Areas: NOAA Sea Grant Population and Ecosystem Dynamics Fellowship

].pull-right[

]

---

---

# An Optimistic Skeptic

### The skeptic...

Someday I hope for a clear answer...

- But it hasn't happened yet. 
  
**Uncertainty and complexity are the norm**

### The optimism...

- There are lessons in null-results

- Slow progress towards better understanding and better outcomes

**Science can help us navigate uncertainty and complexity**

???

---

# How many fish can we catch?

Suppose we conduct a survey and find that there are 100,000 fish in the population...

--
  
  - Does this represent a large fraction of a small but unfished population?
  
  - A small fraction of a historically massive population?
  
--

Which is right has major implications for management.

To answer this, we use a class of models broadly called *stock assessments*

---

# Improving Estimates of Global Fishery Status

- Many efforts to assess the status of global fisheries 
  - UN Sustainable Development Goals
  
- Nearly all are versions "catch-only" data-limited assessments

].pull-right[

Collage of "status of global unassessed" papers

]

---

# Catch = Fish?

---

# Data Limited Stock Assessments

Most fisheries don't have the data or resources for integrated stock assessment

Growing push for cheaper, simpler, but effective methods

All trade data for assumptions, but vary in their details

- Indicator based
  - Length based
  - Catch based

**How well do they work and how good is good enough?**

---

# Catch often != Fish

For those not impressed by simulations, here are some real world data

---

# Inside vs. Outside: Channel Islands

---

# Forecasting Salmon by Age Group

---

# Forecasting with Ensembles

---

# Innovations in Assessment and Management

.center[![:scale 80%](data:image/png;base64,#https://danovando.github.io/dissertation/thesis_files/figure-html/cs2-1.png)]

---

# Code is a Tool.

- Just because we can code it doesn't mean it's right

- Coding is easy (relatively speaking).

- Ideas are hard
  - ideas should drive code, not *vice versa*
  
- We need to be willing to put our computer games to the test

].pull-right[

.center[![](data:image/png;base64,#https://gmedia.playstation.com/is/image/SIEPDC/horizon-forbidden-west-screenshot-01-disclaimer-02oct20?$2400px$)]

]

---

# Inside vs. Outside = Effect?

---

# Inside vs. Outside = Effect?

.center[![:scale 80%](data:image/png;base64,#https://raw.githubusercontent.com/DanOvando/weird-fishes/master/static/blog/2021-07-06-pop-effects-of-mpas_files/biased_dr_plot.jpg)]

---
# Anatomy of a Catch-Only Model

Given a catch history and a population model...

1. Construct **prior distributions** of parameters
  - growth rate, carrying capacity, initial and final stock status
  
2. Sample parameters from prior distributions and apply to catch history + model

3. **reject** combinations of parameters that crash the population or **don't conform to status priors**

].pull-right[

]

---

# Simulating Social-Ecological Systems with  [`marlin`](https://danovando.github.io/marlin/)

.center[![:scale 80%](data:image/png;base64,#https://danovando.github.io/marlin/reference/figures/README-unnamed-chunk-10-1.png)]

---

# Simulating Social-Ecological Systems with  [`marlin`](https://danovando.github.io/marlin/)

---

# What Is an MPA?

---

# What Is an MPA?
.center[![:scale 80%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/mpa1.jpg)]
---

# Catch-Only `sraplus`

A good model should be able to **learn** something about stock status when confronted with **data**

.center[![:scale 75%](data:image/png;base64,#/Users/danovan/teaching/osu-seminar/imgs/catch_status_plot.png)]

---

# Science of Protected Areas

Protected areas asked to play a growing role in **ecosystem-based management**
  
  - 30% of land and water in the coming decade.
  
  - Up from single digits in global seas now

Achieving *ecological* and *social* benefits through protected areas complicated by

- Spatial-temporal dynamics
  
  - Species interactions
  
  - Human behavior

???
Makes *predicting* and *assessing* the performance of protected areas harder than you might think.

[Currently 26% MPA in US waters, 3% no-take MPAs](https://marineprotectedareas.noaa.gov/gallery/understanding-area-based-mgmt-in-us-waters.html)
---

# Social-Ecological Impacts of Protected Areas

- The bigger the protected area...
  - The more likely substantial conservation success
  - The less likely benefits to fisheries

- Surprisingly limited empirical evidence

- Goal at OSU/ODFW: **Empirical research on the impacts of MPAs**
  - Focus on Oregon & California Current
  - Applications globally

]
.pull-right[

]

---

# Social-Ecological Impacts of Protected Areas

.center[![:scale 100%](data:image/png;base64,#https://danovando.github.io/dissertation/thesis_files/figure-html/catch-effects-1.png)]

---

# How Many Fish Are There?

> Counting fish is just like counting trees, except they move and are invisible
> <footer>---  John Shepherd (loosely)</footer>

Two general options

1. Conduct fishery-independent surveys
    
    - Research trawls, acoustics, visual surveys, etc.
    
    - Accurate (hopefully) but expensive
    
  2. Depend on data from the fishery
   
    - Fisheries captures
   
    - Catch-per-unit-effort
   
    - Cheaper but often biased

---

# How Many Can we Catch?

To contextualize numbers, we fit **population dynamics** models
    
  - Births
  
  - Deaths
  
  - Growth

using ideally fishery **dependent** and **independent** data

These models estimate **stock status** relative to **reference points**
  
  - Unfished biomass
  
  - Biomass that would maximize long-term catch
  
  - Sustainability of fishing effort

---